On 5/30/2026 2:32 AM, Abhigyan Kumar wrote:
> sstc function incorrectly checks for avaibility of env->rdtime_fn.
> This causes it to fail each time it's called because rdtime_fn is setup
> later in the procedure (when timer devices are created). This prevents
> stimecmp's addition to gdb.
>
> This change ensures only the riscv_cpu_cfg(env) check is done. rdtime_fn
> will always be NULL otherwise and fail.
I am not sure we can do that - there are several boards that don't create
the timer that implements rdtime_fn (e.g. spike). If we just remove the
check we'll break the sstc check for them.
I suggest using the 'env->debugger' flag to skip the rdtimer check if
we're running the gdbstub:
$ git diff
diff --git a/target/riscv/csr.c b/target/riscv/csr.c
index 5514e0f455..63f4795ac7 100644
--- a/target/riscv/csr.c
+++ b/target/riscv/csr.c
@@ -590,7 +590,8 @@ static RISCVException sstc(CPURISCVState *env, int csrno)
{
bool hmode_check = false;
- if (!riscv_cpu_cfg(env)->ext_sstc || !env->rdtime_fn) {
+ if (!riscv_cpu_cfg(env)->ext_sstc
+ || (!env->debugger && !env->rdtime_fn)) {
return RISCV_EXCP_ILLEGAL_INST;
}
Testing with gitlab 3496 this seems to work:
$ riscv64-unknown-elf-gdb
(gdb) target extended-remote :1234
(gdb) p $stimecmp
$1 = 0
(gdb)
Thanks,
Daniel
>
> Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3496
> Signed-off-by: Abhigyan Kumar <314abh@gmail.com>
> ---
> target/riscv/csr.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/target/riscv/csr.c b/target/riscv/csr.c
> index f5b0895fd..d72df945f 100644
> --- a/target/riscv/csr.c
> +++ b/target/riscv/csr.c
> @@ -590,7 +590,7 @@ static RISCVException sstc(CPURISCVState *env, int csrno)
> {
> bool hmode_check = false;
>
> - if (!riscv_cpu_cfg(env)->ext_sstc || !env->rdtime_fn) {
> + if (!riscv_cpu_cfg(env)->ext_sstc) {
> return RISCV_EXCP_ILLEGAL_INST;
> }
>